🐛 fix(fuzz): size harness buffers to their input

[gaborbernat]

The IDNA and phone harnesses decoded UTF-8 into a block with one slot per input byte. A multibyte input decodes to fewer code points than bytes, so the engines got a buffer with spare slots past the last code point, and ASan missed a read into them. All three standalone harnesses, the JS one included, also gave an empty input one slot.

Each harness now counts the code points first and allocates that many, none for an empty input. The file readers allocate the file size, none for an empty file. libFuzzer copies each input into a heap block of Size bytes for the same reason, so ASan catches an overflow of it (FuzzerLoop.cpp).

With a harness patched to read one code point past the end, the old allocation let a multibyte and an empty input pass clean, and the new one stops both with a heap-buffer-overflow report. The fixed harnesses run the seed corpora and tox -e asan-js clean under ASan and UBSan.

[codspeed]

Merging this PR will regress 1 benchmark

⚡ 1 improved benchmark
❌ 1 regressed benchmark
✅ 579 untouched benchmarks
⏩ 32 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
❌	test_feature[prune-shared-single]	89.1 µs	96.7 µs	-7.88%
⚡	test_feature[shadow-slot-comments]	137.9 µs	110.1 µs	+25.23%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing gaborbernat:fix/harness-exact-buffers (8a955de) with main (af31f20)}

Footnotes

1. 32 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩